Table 2.

Comparison of simulation results

	In silico genotyping			Power (%)
Family structures, with selected individuals genotyped for candidate SNPs	Missing genotypes (% of total)	Percentage of missing genotypes inferred	Error rate (%)	Simulateda H2	Using observed genotypes	Using inferred genotypes	After genotyping all individuals
CEPH pedigrees in Table 1	46.4	82.4	0.4	0.5	0.0	83.9	97.1
Ten three-generation pedigrees with eight offspring each
Grandparents and parents genotyped	57.1	82.4	0.4	0.5	0.0	74.1	92.2
Grandparents, parents and one child genotyped	50.0	99.2	0.7	0.5	0.0	91.2	92.2
100 three-generation pedigrees with eight offspring each
Grandparents and parents genotyped	57.1	82.4	0.4	0.05	0.3	44.3	82.2
Grandparents, parents and one child genotyped	50.0	99.2	0.7	0.05	1.3	70.8	82.2
500 nuclear families with three offspring each
Parents and one child genotyped	40.0	94.6	1.1	0.05	4.7	62.1	82.7
Parents and two children genotyped	20.0	96.9	0.6	0.05	40.8	77.4	82.7
500 nuclear families with two offspring each
Parents and one child genotyped	25.0	94.7	1.0	0.05	4.5	27.3	41.2

^aWe simulated genotypes for eight SNPs within an 8-cM region and a trait-determining SNP with a minor allele frequency of 0.3 that explained either 50% or 5% (H²) of the total phenotypic variance. The associated SNP was placed between the 4^th and the 5^th SNP (for details, see Supplementary Methods).